On Thu, Oct 31, 2002 at 05:15:35PM -0500, Spam wrote: # Way down on my Razor wish list, I would like to be able to exclude # file types from Razor scanning, such as .jpg and .gif. I am finding # I get a lot of false positives from people that use little artsy # borders in their messages. Those artsy borders are Razored. A lot # of personal correspondence gets Razored when people use animated # gifs, such as waving dogs or whatever, and those gifs are Razored. # I guess I could get into a revoke war with the people submitting # them, but I don't have a good web interface to revoke a message # right now.
We've had this request a few times now, both directly and indirectly (I want the ability to whitelist/blacklist thing X). The problem with such mechanisms are that they generally work against the system. Razor2/SpamNet has the concept of contention, or contended content classes, which represents significant trusted disagreement over a particular piece of content (in your example, a common image). Once a class becomes contended, it can no longer show up as a SPAM class. If people whitelist and/or blacklist content types, favorite email addresses (like mailing lists), etc, they rob the system of valuable feedback that would otherwise result in a correct determination. Bad or Lazy Person 1 reports mailing list mail, but Otherwise Smart Person 2 wouldn't revoke it because he whitelisted it, thereby causing the system to take longer to achieve a social majority determination of a particular content class' disposition. That said, obviously the goal of any solution should be to present an effective, convenient solution to any user. Putting the whitelist stuff into the SpamNet and Razor2 clients is already pushing the edge of the philosophy above, but adding more may go too far.. It's really difficult to strike a balance between technical convenience and giving the system the input it needs to operate properly.. thoughts? --jordan
msg00285/pgp00000.pgp
Description: PGP signature
